sys: add uses of driver_filter_t where appropriate

[ehem_freebsd_m5p.com]

Summary

These declaractions need to match driver_filter_t. While there are
plenty of other spots where driver_filter_t would be appropriate, these
seem to have distinct value.

[ehem_freebsd_m5p.com]

Simply chasing down some instances where device interrupt filters really should have been using driver_filter_t and weren't. Rather distinct sort of cleanup. I'm unsure of reviewers for this...

[mhorne]

Seems fine to me. Is this exhaustive or is it just the drivers you encountered?

[jhb]

We don't generally require the typedefs to be used for prototypes. For example, very few device driver "normal" interrupt handlers use driver_intr_t. There is some mix in the tree of using the typedefs for cdevsw methods (some drivers use the typedefs, some use "plain" prototypes). However, for most of the new-bus related functions typedefs are fairly rare in prototypes.

[ehem_freebsd_m5p.com]

These are merely some of the ones I've identified. Notably these were ones where this seemed the right approach, as others would need more adjustment to make them consistent.

[ehem_freebsd_m5p.com]

I've got a patch with the rest of the list. There are around 3-5x this batch for the full list. I could merge that into D32345, or I could create Phabricator entry for those. The majority are various NICs, but there are a few non-NICs and a some architecture functions.

A few functions implicitly marked global stood out as being likely to be redeclared static.